[CLSTR]
< Back to situations

Monitor this situation.

[SITUATION] · [QUIET] · [TECHNOLOGY]

2 clusters · 6 sources · 3 days · First seen · Last updated

Ataraxos AI development for Stratego

Overview

Researchers from MIT, Carnegie Mellon, New York University, and Stanford have developed Ataraxos, an artificial intelligence system designed to master the strategy game Stratego. The system is notable for its ability to handle games of imperfect information, where opponent piece identities are hidden.

Ataraxos has demonstrated high-level proficiency by defeating top human players, including four-time world champion Pim Niemeijer, in a series of matches. To manage the game’s massive search space, the AI utilizes a dual neural network architecture. This includes a generative belief model that estimates the identities of hidden pieces based on movement patterns, allowing the system to sample plausible game states.

The AI was trained through self-play reinforcement learning, completing approximately 163 million games. Compared to previous high-cost models, Ataraxos was developed with high efficiency, requiring significantly less data and fewer computational resources, such as only 16 GPUs for one week of training.

Entities

Stanford University · Carnegie Mellon University · Ataraxos · MIT · Pim Niemeijer

Timeline

  1. [TECHNOLOGY] 4 sources
    Ataraxos AI masters Stratego by predicting hidden pieces

    Researchers from several major universities developed Ataraxos, an AI that mastered Stratego by using a belief model to predict hidden enemy pieces, defeating world champion Pim Niemeijer.

  2. [TECHNOLOGY] 2 sources
    Ataraxos AI defeats top human players in Stratego

    An AI system named Ataraxos has defeated top human players in the strategy game Stratego, demonstrating high efficiency in mastering games of imperfect information.

Sources

apple.slashdot.org · bright.nl · dutchcowboys.nl · elburgado.com · gry-online.pl · nature.com